Methodology of hybrid information support for making optimal decisions on managing dynamic processes: neural network and quantum computing support

Classical information support for optimal control of dynamic processes is a set of methods, models, and computational tools that enable the collection, coordination, evaluation, forecasting, and interpretation of data on the state of an object in order to formulate an acceptable and optimal control action under conditions of uncertainty, constraints, and multicriteria. However, in today’s environment of sensor saturation and digitalization of technological processes, fundamental difficulties arise associated with the exponential growth of the computational complexity of algorithms for synthesizing optimal solutions as the dimensionality of the state space increases, environmental uncertainty increases, and a priori information on the current state of multidimensional processes/control objects is incomplete/noisy. The aim of this work is to propose a concept for information support for optimal decision-making in the control of dynamic processes in high-dimensional problems, based on the rejection of the deterministic calculation of a single optimal solution. Instead, a hybrid information space is formed where neural network operators perform continuous dimensionality reduction and fast feedback approximation, and quantum processors are used as specialized accelerators for two NP-hard subproblems. An example of implementing a dynamic neural network for approximate solutions at the second stage of the hybrid information system’s operational sequence is presented. A procedure for quadratic discrete quantum optimization without constraints based on adiabatic quantum annealing is described, applicable at the final stage of the search for an optimal solution.